From: Stephen D. Williams (sdw@lig.net)
Date: Tue May 08 2001 - 12:17:15 PDT
Brian Atkins wrote:
>
> (when did Netscape stop reminding me that I was about to send a blank?)
>
> Does anyone know what OS IBM will be running on those little Blue Gene
> CPUs?
Couldn't find out, but a competitor is:
http://news.cnet.com/news/0-1003-200-1425245.html?tag=rltdnws
In any case, with a massively scalable CPU/memory architecture, you
wouldn't really need the full operating system on each processor. In
other words, it's non-SMP by definition. You run the full OS on a
different type of processor or a small subset and matrix processes on
the work units.
You guys may have missed it, but a couple of weeks ago a guy from IBM
Germany submitted some patches to the Linux kernel that removed the Hz
clock tick in favor of scheduled timer interrupts because when he ran
1000+ Linux virtual machines on the 370, the 100Hz timer tick was using
a lot of CPU. Obviously there's a good chance that all that 370
hardware may have interesting new life! Of course this also helps UML
(User Mode Linux) which is pretty nifty.
sdw
>
> Eugene Leitl wrote:
> >
> > On Tue, 8 May 2001, David Stutz wrote:
> >
> > > I'm just a doddering old fool to think that the local OS matters at all,
> >
> > Of course a local OS matters. If my desktop box has a 1 k CPUs,
> > they can't be fat CPUs, nor can they have large memory grains. So the
> > amount of shared code present in each node better be minimal.
> > I can't afford to replicate the bloat all over the place.
> >
> > So, some OSses can go places other OSses can't.
> >
> > [eleitl@pc27 DSPsrc]$ ls -l /boot/vmlinuz-2.2.17-21mdk
> > -rw-r--r-- 1 root root 665029 Oct 5 2000 /boot/vmlinuz-2.2.17-21mdk
> >
> > Excuse me while I bust an aneurism laughing. How is that bloated thing
> > supposed to scale to a 1 MBit grain architecture?
>
> --
> Brian Atkins
> Director, Singularity Institute for Artificial Intelligence
> http://www.singinst.org/
-- sdw@lig.net http://sdw.st Stephen D. Williams 43392 Wayside Cir,Ashburn,VA 20147-4622 703-724-0118W 703-995-0407Fax Dec2000
This archive was generated by hypermail 2b29 : Tue May 08 2001 - 12:28:26 PDT